GSI/WRF Regional Data Assimilation System and Its Application in theWeather Forecasts over Southwest Asia - Data Assimilation for Atmospheric, Oceanic and Hydrologic Applications - page 627

diurnal cycle of surface air temperature can be improved slightly by the assimilation
of satellite radiance data.
For the analysis of the 10-m wind fields, the reference data used is still the NCEP
GFS analysis data. In contrast to surface temperature, it is not readily apparent that
the amplitude of the diurnal cycle has been improved in DA (Fig. 25.18 b). The
performance is quite different in these selected sub-regions. The diurnal cycle of
the wind speed (Fig. 25.18 b) in the analysis data is considerably larger than in the
model forecasts over the five sub-regions B, D, E, H, and I, where B, D and E are
three high mountain sub-regions. But it is clear that the amplitude of the diurnal
cycle in the DA experiment has been modified closer to the analysis data.

25.7 Summary and Discussion

25.7.1 Summary

This paper presented an objective verification and impact of radiance data assimila-
tion on weather forecasts over the complex terrain areas of Southwest Asia using the
National Center for Atmospheric Research (NCAR) mesoscale model (WRF-ARW)
and Joint Center for Satellite Data Assimilation (JCSDA) GSI analysis system. The
numerical experiments are conducted for a one month period May 2006. The results
are summarized as follows:
The model biases caused by inadequate parameterization of physical processes,
except for the 2-m temperature, are relatively small compared to the nonsystematic
errors resulting, in part, from the uncertainty in the initial conditions. The total
forecast errors at the surface show a substantial spatial heterogeneity; there is a
relatively larger error in the higher mountain areas. However, the sources of the error
indicate a unique difference between temperature, precipitation and wind speed.
While the error in 2-m temperature is mainly from systematic error, which is largely
controlled by the physical representation of terrain (i.e., the errors are positively
correlated with terrain elevation); the errors in 10-m wind speed and precipitation
have a greater contribution from nonsystematic error, which is more likely related
to uncertainty in the initial conditions.
The amplitude of the diurnal cycle of the model 2-m temperature is much smaller
than the GTS observations. However, the model forecasts of the diurnal cycle are
consistent with the NCEP GFS analysis data. There is no evidence of a sharp gap
between the model forecasts and the analysis data.
The ATOVS satellite data provides useful information for improving the initial
conditions, and the model error was reduced to some degree. The bias and mean
square error skill score (SS) shows that satellite data assimilation produces a better
forecast over some areas; however, it seems not to make a significant contribution to
the accuracy of forecasts in the higher mountain areas. Although the improvement
in correlation coefficient growth is very small, the forecast patterns are improved in
the DA experiment,.